chore: publish benchmarking results in json file (#3110)

Signed-off-by: Sahithi Chigurupati <[email protected]>

Author: csahithi

components/backends/trtllm/performance_sweeps/scripts/bench.sh

@@ -163,8 +163,10 @@ python3 ${SCRIPTS_DIR}/scripts/bench/benchmark_serving.py \
         --host ${hostname} \
         --port ${port}
 
+mkdir -p ${log_path}/results
 echo "Starting benchmark..."
 for concurrency in ${concurrency_list}; do
+    original_concurrency=${concurrency}
     concurrency=$((concurrency * num_gen_servers))
     num_prompts=$((concurrency * multi_round))
     echo "Benchmarking with concurrency ${concurrency} ... ${num_prompts} prompts"
@@ -186,7 +188,10 @@ for concurrency in ${concurrency_list}; do
         --percentile-metrics ttft,tpot,itl,e2el \
         --max-concurrency "$concurrency" \
         --host ${hostname} \
-        --port ${port}
+        --port ${port} \
+        --save-result \
+        --result-dir "${log_path}/results" \
+        --result-filename "results_concurrency_${original_concurrency}_gpus_${total_gpus}.json"
 
     echo "Benchmark with concurrency ${concurrency} done"
     do_get_logs ${log_path}/output_workers.log ${log_path}/concurrency_${concurrency}